Why SpaceTravLR 🌔️ ?

SpaceTravLR (Spatially perturbing Transcription factors, Ligands & Receptors)

SpaceTravLR leverages convolutional neural networks to generate a sparse graph with differentiable edges. This enables signals to propagate both within cells through regulatory edges and between cells through ligand–mediated connections.

Core Features

• predicting niche-specific perturbation outcome at single cell resolution
• inferring functional cell-cell communications events
• identifying spatial domains and functional microniches and their driver genes

Quick start

Make & sync your Environment the modern way

pip install -r requirements.txt

uv pip install SpaceTravLR==0.1.17

Installing from Source

uv venv
source .venv/bin/activate
uv sync

Load the example Slide-tags Human Tonsil data.

adata = sc.read_h5ad('data/snrna_germinal_center.h5ad')

Create a SpaceShip

from SpaceTravLR.spaceship import SpaceShip

spacetravlr = SpaceShip(name='myTonsil').setup_(adata)

assert spacetravlr.is_everything_ok()

spacetravlr.spawn_worker(
    python_path='.venv/bin/python',
    partition='preempt'
)

SpaceTravLR generates a queue of genes that each worker consumes in parallel. spacetravlr.spawn_worker submits a new job to the clusters.

Outputs

output/
├── input_data/
│   ├── _adata.h5ad
│   ├── celloracle_links.pkl
│   ├── communication.pkl
│   ├── LRs.parquet
├── betadata/
│   ├── PAX5_betadata.parquet
│   ├── FOXO1_betadata.parquet
│   ├── CD79A_betadata.parquet
│   ├── ...
│   ├── IL21_betadata.parquet
│   ├── IL4_betadata.parquet
│   ├── CCR4_betadata.parquet
├── logs/
│   ├── training_TIMESTAMP.log

Results

Citation

If you find SpaceTravLR useful in your research or projects, please cite our paper: